Joel Nothman

> What I've called a "release" (but we should maybe rename) is any visible dataset, so there's always a review stage before visibility. Yes, and we could still have logic...

Think of the v1, v2, v3, more like Git's commit ref SHAs.

First step is to explore our uploader code (i.e. each step) to see what needs to be done to adapt it to the update use case.

> Do we create an upload dir for editing the existing dataset and copy data from previous repo dir If we will not allow multiple concurrent edits to a dataset,...

An extra note re > When user clicks update the dataset, frontend will send the dataset id to backend and get back a bunch of data, including the upload type,...

Need to think about what to do with checking for duplicate images... will need to exclude matches against the upload that is being updated..?

I've tried to think about the transitions we care about when we allow drafts. [Editable graph here](https://dreampuf.github.io/GraphvizOnline/#digraph%20G%20%7B%0A%20%20NewUpload%20-%3E%20AwaitingReview%20%5Blabel%3D%22Submit%22%5D%0A%20%20AwaitingReview%20-%3E%20DraftCompleted%20%5Blabel%3D%22Approve%20%5Bdraft%5D%22%5D%0A%20%20AwaitingReview%20-%3E%20ReleaseCompleted%20%5Blabel%3D%22Approve%20%5Brelease%5D%22%5D%0A%20%20AwaitingReview%20-%3E%20Failed%20%5Blabel%3D%22Reject%22%5D%0A%20%20AwaitingReview%20-%3E%20NewUpload%20%5Blabel%3D%22Request%20revisions%20%2F%20Invalid%22%5D%0A%0A%20%20DraftCompleted%20-%3E%20UpdatingDraft%20%5Blabel%3D%22Edit%22%5D%0A%20%20UpdatingDraft%20-%3E%20DraftCompleted%20%5Blabel%3D%22Save%22%5D%0A%20%20UpdatingDraft%20-%3E%20DraftCompleted%20%5Blabel%3D%22Rollback%22%5D%0A%0A%20%20ReleaseCompleted%20-%3E%20UpdatingRelease%20%5Blabel%3D%22Edit%22%5D%0A%20%20UpdatingRelease%20-%3E%20ReleaseCompleted%20%5Blabel%3D%22Save%22%5D%0A%20%20UpdatingRelease%20-%3E%20ReleaseCompleted%20%5Blabel%3D%22Rollback%22%5D%0A%20%20%0A%20%20DraftCompleted%20-%3E%20AwaitingReleaseReview%20%5Blabel%3D%22Release%22%5D%0A%20%20AwaitingReleaseReview%20-%3E%20ReleaseCompleted%20%5Blabel%3D%22Approve%20%5Brelease%5D%22%5D%0A%20%20AwaitingReleaseReview%20-%3E%20DraftCompleted%20%5Blabel%3D%22Reject%22%5D%0A%20%20%0A%20%20DraftCompleted%20-%3E%20Failed%20%5Blabel%3D%22Delete%22%5D%0A%20%20ReleaseCompleted%20-%3E%20Failed%20%5Blabel%3D%22Delete%22%5D%0A%7D)  This suggests that: * for simplicity, we don't allow draft uploads,...

You can ignore the stuff related to drafts there, for now.

Using the same upload ID has some convenience, but my questions surround: * concurrency - if someone has two editors open for the same dataset, how does that work? *...

And maybe even any directories whose files are all untouched in > n days. (or from users whose sessions weren't active in as many days, but this may mess up...